American Journal of Environment and Sustainable Development Vol. 3, No. 3, 2018, pp. 46-54 http://www.aiscience.org/journal/ajesd

Urban Growth and Deforestation by Remote Sensing in the Humid Tropical Forest of Congo Bassin: Case of in Republic of Congo

Stoffenne Binsangou 1, Suspense Averti Ifo 1, *, Lypsia Ibocko 1, Lisa Louvouandou 1, Mesmin Tchindjang 3, Félix Koubouana 2

1Laboratory of Geomatic and Applied Tropical Ecology, Higher Normal School, University of Marien N’GOUABI, , Congo 2Laboratory of Geomatic and Applied Tropical Ecology, Higher National School of Forestry and Agriculture, University of Marien N’GOUABI, Brazzaville, Congo 3Department of Geography, University of Yaoundé I, Yaoundé, Cameroon

Abstract Urban growth is identified as one of the main deforestation drivers and land degradation in tropical region. Very few data exist to quantify the impact of this drivers. The aims of this study were: (i) estimating the activity data of the growth of the city’s Impfondo in the period between 1986 to 2016; (ii) determining the indirect drivers of that urban growth. This study is carried out in Impfondo. The remote sensing and GIS techniques were used to assess urban growth and its impact on deforestation and degradation of the forest around and in Impfondo the period between 1986 and 2016. Landsat images were used; specially 2 images Thematic Mapper (TM) of 1986 and Operational Land Imager (OLI) of 2016. ERDAS 2014 software was for images classification, classification and calculation of forest lost area during this growth. Additional field survey was made to collect truth GPS ground points. As result, the urban tram area growth from 267.86 ha in the year 1986 to 859.68 ha in the year 2016. This study showed the impact of urban growth on deforestation on the one hand and on the other hand stress considering this factor in mitigating the effects of climate change.

Keywords Forests, Deforestation, Remote Sensing, Urban Growth, Impfondo,

Received: August 19, 2018 / Accepted: September 11, 2018 / Published online: October 9, 2018 @ 2018 The Authors. Published by American Institute of Science. This Open Access article is under the CC BY license. http://creativecommons.org/licenses/by/4.0/

sustainability of tropical forest ecosystem. Although some 1. Introduction authors [5, 4] consider that the Congo basin forest are relatively well preserved, the deforestation rate (0.17%) is far The forests of the Congo Basin extend over nearly 200 below in comparison with one observed in the others tropical million hectares and 90% of the forest consists of dense regions e.g Amazonian forest. The rate of deforestation in tropical forests. More than 99% of the forest surface consists Amazonian forest reached 0.4% per year annum [6]. in the of primary or naturally regenerated forests, as opposed to past decades. Deforestation and degradation have significant plantations. These tropical forests are characterized by impacts on plant biodiversity [1] and contribute significant biodiversity [1, 2, 3]. However, these forests approximately to almost 18% of global greenhouse gas suffer from both quantitative and qualitative loss of emissions [7-10]. Several authors have reviewed the causes biodiversity resulting. However, these forests suffer a loss of deforestation and degradation in the tropics [11, 12], and both quantitatively and qualitatively because of deforestation shown that agriculture, logging activities were identified as and forest degradation [4]. with significant impacts on the

* Corresponding author E-mail address: 47 Stoffenne Binsangou et al. : Urban Growth and Deforestation by Remote Sensing in the Humid Tropical Forest of Congo Bassin Between 1986–2016: Case of Impfondo, Republic of Congo the main drivers. from neighbouring countries, such is the case of the urban In the Republic of Congo (RoC), the annual deforestation community of Impfondo. rate growth from 0.02% in 1990-2000 to 0.03% in 2000-2010 The Impfondo city was created on December 12 th , 1920 and [13]. This deforestation is unevenly distributed throughout is the head quarter of the Likouala region. Historical data the national territory of RoC. It has been noted in some collected in the field revealed that this urban community was departments of RoC that the rates of deforestation were much originally a camp of fisherman settled on the Ubangi River. higher than the national average rate. For 2000-2010: The In addition, this camp was plunged into a forested area. This department of Kouilou had a deforestation rate of 1.78%, urban community has widened with time attracting the while in the other departments (Lekoumou, Niari and Pool) populations of the villages scattered around. Also, the arrival the rate of deforestation was 1.06%, 1.35%, 0.91% of the first catholic missionaries help in promoting the respectively). Several factors were identified as the source of administrative organization of this locality. forest losses in the Republic of the Congo, including: The present study aimed at quantifying the areas of forest agriculture, infrastructure development, wood energy loss because of the development of this local urban collection and urban growth, etc [24]. Urban growth also community. To date, starting from the official date of the explains the significant losses of forest due to the creation of the Impfondo city, how much was the evolution consumption of building materials and the land use for house of the urban community? What factors contributed to its building and rural activities developments. development? Urban land expansion is placing a formidable challenge in The following assumptions were made: the growth of the many countries around the globe, especially in terms of their urban community of Impfondo and its related would be the link with deforestation and land degradation in intertropical main cause of forest loss in that community; Population region where many cities were developed inner deep tropical growth, migratory population flows would justify this urban [14, 13]. According to a report submitted by the REDD + expansion and thus deforestation in the study area. The MDDEF-GAF pilot project "Monitoring National Forest objectives of this study were to assess the impact of urban cover" in the Republic of the Congo, the second major growth on the forest and to determine its immediate and drivers of Deforestation is the expansion of residential areas structural drivers. (including logging roads), causing 14% of the changes in forest cover in the Republic of the Congo in the period of 2000-2010. This study had shown the bond which existed 2. Materials and Methods between the zones of conflicts and the increase in the of the 2.1. Presentation of the Study Area population in the localities having which had received the refugees of civil war come from bordering countries. Urban The study was carried out within the tropical rainforest of the land expansion a universal phenomenon that has known and North of Congo Brazzaville in the department of Likouala is currently experiencing a acceleration in Africa and mainly (Figure 1). The zone of study covers a total surface of of the republic of Congo [15, 05]. As urban growth is 155274 ha. It lies between 1∘ 27’ 52.85’’ and 2∘ 6’55.76’’ of concerned, several factors have been identified: the northern latitude and between 17 ∘ 52’ 35.04’’ and 18 ∘ population increase, rural exodus, civil war and conflicts with 04’32.65’’ of longitude. The climate of the study area was of migratory phenomena accompanying it in some Central equatorial type. Mean rainfall is of 1760 mm y−1 , with a dry Africa countries, Poverty stricken in the rural area, the season from December to January and a long-wet season development of social and economic of cities [16, 17, 13]. from March to November (Figure 2). In the Dongou district, Observations in some countries of the Congo basin revealed the soil cover is of tertiary clay sandy formation and a that African cities are spreading and taking over natural quaternary alluvial formation to the east. spaces. In the Republic of the Congo, a recent study The soils derived from there are impoverished ferrilitic highlights the causes of urban sprawl: population growth, brown-red clay-sand soils on the Western plateau, land-use planning, accelerated municipalization and ferralitic/hydromorphic alluvial soils on alluvial terraces, and residential sprawl have participated to the transformation of waterlogged peat soils in flooded areas. This area has one of the landscape from south of Country until the North of the very low densities of human population (0.93 km2) of the territory. This study also showed how urban sprawl in a Republic of Congo. forest town, the city of Ouesso, has resulted in significant The forest of Likouala contains a high diversity of trees and forest losses [15]. Other regions of the Republic of the Congo plants [1]. In the Dongou district, the forests of the study area experienced in the years 1990 and 2000 to 2010 an important are rainforest. The principal vegetation types are partially development of urban space due to the large flow of refugees American Journal of Environment and Sustainable Development Vol. 3, No. 3, 2018, pp. 46-54 48

deciduous dense rainforests of Ulmaceae and Sterculiaceae, forest of Guibourtia demeusei (Harms) J. Léonard [1]. Tree swampy flooded forest of Uapaca heudelotii Baill. , and canopy closure of the forest varies from 93% to 100%

Figure 1. Locating the study area.

Figure 2. Ombrothermique diagram of the Likouala (mean of 1932-2015), ANAC Congo (2016).

49 Stoffenne Binsangou et al. : Urban Growth and Deforestation by Remote Sensing in the Humid Tropical Forest of Congo Bassin Between 1986–2016: Case of Impfondo, Republic of Congo

2.2. Methods were done; image classification, urban growth/land cover change analysis and deforestation assessment. The following steps were used to achieve the objectives of the study: Landsat images data collection from the website The image of the study area was extracted from the all scene http://glovis.usgs.gov/ (2 scenes 181-59); image using the subset image tool to facilitate this study (Figure 3: enhancement: During the image enhancement step, geometric Color composition 7.4.2). The study area covers an area of and radiometric restitutions of the 2 sets of Landsat images 2940.4 ha.

Figure 3. Landsat OLI Image with 7.4.2 bands.

2.3. Characteristics of the Images Used to applied for the 1986 image and the 7-4-2 bands for the 2016 Carry out This Study image. Two types of images were used for this study: the Landsat Image supervised classification was done using maximum image for 1986 and the OLI image for 2016. These two likelihood algorithm. In the urban growth/land cover change images which had the path and row number 181-059, were analyses phase, post classification comparisons and analyses captured on different dates, in December for image of 1986 were performed to reveal the urban growth/land cover change and February for the image of 2016, having both a resolution pattern in the study area between the two dates. As well as of 30m. The color composition of the 5-4-3 bands was deforestation assessment is concerned, image re- classification and cross-tabulation analysis were performed to American Journal of Environment and Sustainable Development Vol. 3, No. 3, 2018, pp. 46-54 50

assess the extent of deforestation in the Impfondo area. Committee - CNAR); on accelerated Municipalization (a The bands selection was made by considering the bands that government program intends to develop the administrative structures of cities by building infrastructure) in the Likouala allow us to discriminate urban plot, forest and non-forest Prefecture; on the causes of deforestation from local officials vegetation. The pretreatment of Images proceeded by a and authorities. separate band visualization of the study area to determine which bands were reflective of the development information of the urban area compared to the vegetation. For Thematic 3. Results Mapper (TM) of 1986, we used the colored composition 5-4- 3; While for Operational Land Imager (OLI) of 2016, we 3.1. Occupation of the Soil in the Study used the colored composition 7-4-2. Area After obtaining the first map of land cover type of the study Land cover type of the study area shows us that forest is the area, a field visit was carried out to correct the errors due to most important class in 1986. Overall classification accuracy the confusion of the pixels during the selection of the region is 86.84%. Producer’s accuracy of urban plot class is 98, of interest in laboratory. For this, 100 checkpoints scattered 08% and non-forest class is 62.50%. over the entire map produced had been selected in the The land use map of the study area in 2016 shows an extent laboratory, saved in the GPS map 60CSx of Garmin to of each class in the study area between the two periods perform class checking on the using of Garmin tools. To (Figure 4). In 1986, the forest class was more important than assess the accuracy of the map products, the confusion matrix the other two classes: the urban plot class and the non-forest was produced, and the overall classification accuracy was class. In 1986, urban plot accounted for 9.11% compared to calculated. 29.23% in 2016. Forest accounts for 74% of the study area in 1986, while it increases to 42.74% in 2016. We also noticed 2.4. Field Data Collection an increase of 16.23% to 28.01% between 1986 and 2016 of Other data were collected as part of this study on the ground non-forest item (Table 1). towards the local authorities of the administrative entities, Producers accuracy of the different land use type were of data: on the demographics characteristics of this urban 75.00% for urban plot, 94.44% for the forest and of 96.43% community from 1986 to 2016; on Likouala population for the non-forest when users’ accuracy was 100%, 94.44% poverty; on civil war refugees from neighbouring countries and 87.10% respectively. (obtained from to the National Refugee Assistance

Figure 4. Land use maps in 1986 (left) and 2016 (right). 51 Stoffenne Binsangou et al. : Urban Growth and Deforestation by Remote Sensing in the Humid Tropical Forest of Congo Bassin Between 1986–2016: Case of Impfondo, Republic of Congo

Figure 5. Land cover conversion between 1986 and 2016.

Table 1. Area of the different classes of land use between 1986 and 2016. tram has a direct impact on the sustainability of the dense tropical forest of this region. 1986 2016 Land use Area (ha) % Area (ha) % Urban growth has been identified as a direct cause of Uban plot 267.86133 9.11 859.68 29.23 deforestation by many researchers [5, 4]. As we observed in Non-forest 477.21069 16.23 823.68 28.01 Forest 2195.22798 74.66 1256.94 42.74 the land use map, a large forest area has been replaced by non-forest land with the development of the Impfondo urban 3.2. Deforestation and Urban Growth plot from 1986 to 2016. During this period of study, the At a time, difference of 30 years, urban growth caused a urban tram area growth from 267.86 ha to 859.68 ha a loss of forest loss of 592 ha. The results obtained shows that the the forest area of 591.82 ha. Several factors could explain Impfondo city is sprawling Eastwards (river Ubangi with this urban growth in the municipality of Impfondo: flooded forest) to Westwards (remained dense tropical forest population growth, residential sprawl, land-use planning and ecosystem). the development of peri urban agriculture. These same factors were identified by several authors in the literature 3.3. Drivers of Deforestation having worked in other cities around the world [14, 13, 19, 20] Data collected in the field revealed that both direct and indirect drivers explains the deforestation around the 4.1. Population Explosion Impfondo city. The main cause of deforestation is urban growth, followed by agriculture, new house buildings. The According to the Congo report [5], urban expansion is mainly indirect causes of deforestation in the study area are motivated by the population explosion in large cities. Factors population growth, poverty, migration of people in the last 15 such as rural exodus, economic reasons, displacement and the years coming from Democratic Republic of Congo (DRC), arrival of refugees have been identified as structural causes Rwanda and Republic (CAR). that would explain population expansion in different communities both forested and non-forested. Looking at the results of data collected during field truth ground, the survey 4. Discussion with the local administrative authorities of the urban community of Impfondo revealed that the “accelerated Urban growth is the direct cause of deforestation and municipalization,” the construction of the modern degradation of forest in the North of the republic of Congo. administrative infrastructures in this city in 2004 and 2005 This present study has shown that the development of urban American Journal of Environment and Sustainable Development Vol. 3, No. 3, 2018, pp. 46-54 52

did not contribute significantly to the loss of the forests, that in the Congo Basin, agriculture will be the first cause of because it took place on mostly marshy areas while our changes in forest cover observed. Geist and Lambin [11] research concerns already-subdivided areas. Thus, identified the expansion of agriculture as one of the major infrastructure construction has not been undertaken on groups responsible for deforestation and forest degradation in primary forest. the intertropical zone. Similarly, the REDD + MDDEF-GAF However, an indirect cause of deforestation played a decisive pilot project "Monitoring National Forest cover" in the role in the losses of forests not only in this city, but also Republic of the Congo has come to the results that agricultural expansion has been the main driver of change throughout the Likouala’ region. The massive arrival of representing 78% of changes in forest cover observed in the refugees from neighbouring countries (CAR, DRC and RoC [22]. This observation was confirmed by [24] who Rwanda) greatly participated to the degradation and then the deforestation along the Ubangi river [21]. Indeed, in showed that peasant farming is the main cause of Impfondo in the recent year NGO Médecin d’Afrique have deforestation in tropical areas. It contributes 35% to the destruction of the forest in Africa. registered a total of 30600 refugees from the DRC. This was confirmed by previous studies which showed that from 2000 The importance of agricultural areas around this city could be to 2010, large areas of forest were lost around the towns of explained by the high population density and the increase Betou, Dongou and Impfondo, thus increasing the spatial size demands for food as well as by the level of poverty. of these cities [22]. Agriculture comes in support of the fishery that is exercised The displacement of civil war refugees is an indirect factor by the local people and participates very weakly in regulating for deforestation and degradation of forest in the Republic of the local economy. It is indeed totally dependent on the collection of natural resources from forests, fishing, but also Congo, particularly in the north of RoC, in the Likouala and above all from the sale of basic agricultural products region [15, 22]. such as cassava leaves, oil palm and the development of the Data collected from the government department dealing with banana plantation. When these farmlands became poor, they the management of Refugees (CNAR) reported that from are abandoned for almost seven years for fallow sand such a 2000 to 2014, 13809 refugees had been registered in their situation could explain the large fallows around this forested Impfondo’ office. These people are embarked on agricultural community. activities significantly increasing the useful area of the urban community of Impfondo. Refugees developed the activities in the sector of agricultural inside the primary forest with a 5. Conclusion great impact on the structure and biodiversity of forest. The study on deforestation due to urban growth in the Data on the population collected at National Institute of Impfondo city using remote sensing data between 1986 and Statistics, revealed that the population of Impfondo has risen 2016, reveals that urban growth in forest areas is an from 33.911 to 40.389 habitants from 2007 to 2014. important cause of deforestation and forest degradation in the RoC. The huge amount of forest loss obtained in this study As the population grows, demands for agricultural products allows us to consider urban sprawl as a threat on forest cover increases as well as agricultural production. Around areas. Several causes have led to this situation: population Impfondo a large part of the forests has been destroyed by growth, local spatial planning and the arrival of refugees local populations for the establishment of diverse crops such from neighbouring countries. Nevertheless, it is necessary to as: cassava, banana, maize, peanut, cocoyam and many extend the study to other small size town of RoC. others. These crops respond to a variety of demands from urban populations. But these activities are carried out by the populations not only for their food, but also for the income Acknowledgements they serve to satisfy their needs; but this activity contributes enormously to the growth of this cities. The link between The authors are thankful to OSFACO for funding this project. urban growth and population growth is confirmed by the Special thanks Benoit MERTENS and Camille PINET for numerous authors [19, 8, 22, 17, 13] their kindly support during the time of the project. Authors thanks also our ground staff in Impfondo, Dongou and 4.2. Agriculture Bodzala for their help during the field trip visit. Data collected during fieldwork revealed that agriculture is Research highlights the primary cause of deforestation. Around the Impfondo 1 the urban tram area growth from 267.86 ha in the year urban community, traditional or subsistence agriculture is the 1986 to 859.68 ha in the year 2016, main cause of deforestation. Many studies have concluded 53 Stoffenne Binsangou et al. : Urban Growth and Deforestation by Remote Sensing in the Humid Tropical Forest of Congo Bassin Between 1986–2016: Case of Impfondo, Republic of Congo

2 direct link between human activities and the lost of forest déforestation dans le bassin du Congo. Réconcilier la cover around this agglomeration, croissance économique et la protection de la forêt . International Bank for Reconstruction and Development / the 3 population explosion and poverty are the drivers of world Bank 1818 Hstreet NW, Washington DC 20433, 201p. deforestation [9] Molto Q., (2012), Estimation de la biomasse en forêt tropicale humide: Propagation des incertitudes dans la modélisation de la distribution spatiale de la biomasse en Guyane française, Author’s Contribution Mise: hydrologie et biologie des organismes, populations, interactions, Université des Antilles et de la Guyane , 184p. Suspense Averti IFO, conceived and planned the [10] Niveau des Emissions de Références pour les Forêts (NERF) experiments. Stoffenne BINSANGOU, Suspense Averti IFO, en République du Congo, 2016. Lypsia IBOCKO, Lisa LOUVOUANDOU, carried out the [11] Geist H. J. et Lambin E, F., (2001), III Land-Use and Land- experiments in the field. Ifo Suspense Averti, Stoffenne Cover Change (LUCC) Project IV. International Human BINSANGOU and Mesmin TCHINDJANG contributed to Dimensions Programme on Global Environmental Change the interpretation of the results. Suspense Averti IFO and (IHDP) V. International Geosphere-Biosphere Programme (IGBP), LUCC Report Series. Stoffenne BINSANGOU took the lead in writing the manuscript. Mesmin TCHINDJANG, Félix KOUBOUANA [12] Sahani M., Moeversons J., Vandecasteele I., Ozer P., (2013), provided critical feedback. Expansion urbaine et risques hydrologique: cas de Butembo (RDC), E. planification et aménagement de l’espace. [13] Aguejdad R., (2009), Etalement urbain et évaluation de son Competing Interest impact sur la biodiversité de la reconstitution des trajectoires à la modélisation prospective. Application a une agglomération The authors have declared that no competing interest exists. de taille moyenne: rennes métropole. Thèse pour obtenir le titre de docteur de l’université européenne de Bretagne. Ecole doctorale sciences humaine et sociales, Université renne 2 References haute Bretagnesous le sceau de l’universitéeuropéenne de Bretagne, 378p. [1] Ifo S. A., Moutsambote J-M. Koubouana F., Yoka J., Ndzai S, [14] Binsangou S., Ifo S. A., Koubouana F., Louvouandou L. F., F., Bouetou-Kadilamio L. N. O., Mampouya H. Jourdain C., Nzingoula S. et Mansisse L., (2017), Suivi par télédétection de Bocko Y., Mantota A. B., Mbemba M. Mouanga-Sokath D., la déforestation due à la croissance urbaine de la ville de Odende R. Mondzali L. R., Mampouya Wenina Y. E. Ouissika Ouesso dans la période de 2000–2016, République du Congo. B, C. and Loumeto J. J., (2016), Tree Species Diversity, International journal of Innovation and scientific Research, Richness, and Similarity in Intact and Degraded Forest in the Vol. 30, N°2, Pp. 186-197. Tropical Rainforest of the Congo Basin: Case of the Forest of Likouala in the Republic of Congo. International journal of [15] Champaud J., Chaume R., Cherel J-P., Barret E., Atkinson A. forestery Research , 12p. et Muscat G., (1993), Croissance urbaine, environnement, et imagerie satellitaire, Convention C,E,E, / ORSTOM N° 946 / [2] Puig H., (2001), Diversité spécifique et déforestation: 1990, 295p. l’exemple des forêts tropicales humides du Mexique. Bois For, Trop , 268 (2), 41-55. [16] Kayembe Wa Kayembe M., De Maeyer M. et Wolff E., (2009), Cartographie de la croissance urbaine de Kinshasa [3] Teyssèdre A., (2004), Vers une sixième grande crise (RDC) entre 1995 et 2005 par télédétection satellitaire à haute d’extinctions ? In: Barbault R,&Chevassus-au-Louis B, eds. résolution Belgeo [En ligne], 3-4 | 2009, mis en ligne le 17 Biodiversité et changements globaux, Enjeux de société et mars 2013, consulté le 21 mars 2013, URL: défis pour la recherche. Paris: Association pour la Diffusion http://belgeo,revues,org/7349, pp: 439-455. de la Pensée Française (ADPF) , pp: 24-36. [17] Zhang Y., Guindon B., Sun K., et Sun L., (2012), La [4] Tchatchou B., Sonwa D. J., IFO S. et Tiani A. M,, (2015), télédétection pour améliorer la compréhension de Déforestation et dégradation des forêts dans le bassin du l’urbanisation au canada, Ressources naturelles Canada, Congo: Etat des lieux, causes actuelles et perspectives. Papier Centre Canadien de télédétection , pp: 1-2. occasionnel 120 , Bogor, Indonésie: CIFOR, 47p. [18] Aliate R., (2015). Etalement urbain et deforestation: cas de la [5] Tati G. (1987), La croissance urbaine au Congo (1960-1984). ville d’Ait Melloul. Mémoire d’accomplissement des modules Institut de Formation et de Recherche Démographiques du master «aménagement, développement, gestion et (EFORD) , Yaounde (Cameroun), pp: 177-187. perspective territoriales en zones arides et semi arides. 111p. [6] Demaze M. T., (2007), La déforestation en amazonie [19] Bouassida L., Maréchal C. et Ben Khatra N., (1998), Etude de brésilienne: une rupture apparente entre développement et la dynamique de l’expansion urbaine du grand Tunis (1975- environnement , Le Mans , 23 p 1998). Travaux pratiques, Université d’été du CNES, [7] Lambin E. F. et Geist H. J., (2006), Land-Use and Land-Cover Education Nationale Tunis 07, pp: 1-15. Change: Local Processes and Global Impacts, Springer , [20] Gazel H.; Harre D. et Moriconi-Ebrard F., (2010), Rapport Heidelberg, 222p. général de l’étude «Africapolis II; l’Urbanisation en Afrique [8] Megevand C., Mosnier A., Hourticq J., Sanders K., Centrale et Orientale», Fiche Pays République du Congo, 48P. Doetinchem N. et Streck C., (2013), Dynamique de American Journal of Environment and Sustainable Development Vol. 3, No. 3, 2018, pp. 46-54 54

[21] GAF, (2013), Suivi Forestier Extension des services REDD. [23] Kieffer E. et Serradj A., (2013), La télédétection au service S-Rapport des opérations de service pour la République du des études urbaines: expansion de la ville de Pondichéry entre Congo-phase 3. Doc. n° GSE-REDD-S6-RoC-Ph3, issue/Rev- 1973 et 2009 (Géomatique Expert N° 95), pp: 68-79. n°10, 74p.

[22] Cubry P., (1991), Urbanisation, croissance urbaine et transports en Afrique. Démographe à l’ORSTOM SUPPL au n° 9, Revue générale des chemins de fer , pp: 5